2017 IEEE World Haptics Conference (WHC) Fürstenfeldbruck (Munich), Germany 6–9 June 2017

978-1-5090-1425-5/17/$31.00 ©2017 IEEE

Roughness perception of virtual textures displayed by electrovibrationon touch screens

Yasemin Vardar1∗, Aykut Isleyen1∗, Muhammad K. Saleem1, and Cagatay Basdogan1#

Abstract— In this study, we have investigated the humanroughness perception of periodical textures on an electrostaticdisplay by conducting psychophysical experiments with 10subjects. To generate virtual textures, we used low frequencyunipolar pulse waves in different waveform (sinusoidal, square,saw-tooth, triangle), and spacing. We modulated these waveswith a 3kHz high frequency sinusoidal carrier signal to min-imize perceptional differences due to the electrical filteringof human finger and eliminate low-frequency distortions. Thesubjects were asked to rate 40 different macro textures on aLikert scale of 1-7. We also collected the normal and tangentialforces acting on the fingers of subjects during the experiment.The results of our user study showed that subjects perceivedthe square wave as the roughest while they perceived the otherwaveforms equally rough. The perceived roughness followedan inverted U-shaped curve as a function of groove width,but the peak point shifted to the left compared to the resultsof the earlier studies. Moreover, we found that the roughnessperception of subjects is best correlated with the rate of changeof the contact forces rather than themselves.

I. INTRODUCTION

Touch screens have been used in over a wide range ofportable devices nowadays, but our interactions with thesedevices mainly involve visual and auditory sensory channels,which are already overloaded. While a commercial touchscreen today can easily detect finger position and its gestures,it does not provide the user with haptic feedback. However,haptic feedback can be used in touch screens as an additionalsensory channel to convey information and also reduce theperceptual and cognitive load of the user. For instance, hapticeffects can be used to send a personal message to lovedones in the form of a beating heart, to select the fabric ofa cloth that we purchase online, to teach mathematics toprimary school students (e.g. dragging one number on thesecreen to the other to obtain the desired sum while feelinghaptic resistance for incorrect matching), to differentiate thefeel of riding a bicycle on smooth, bumpy, and sandy roadsduring game playing, and to design a new user interface (e.g.estimating the amount of data in a folder intuitively by thehaptic resistance while dragging it on the screen).

One of the techniques used to generate haptic effects on atouch screen is electrovibration [1]. In this technique, thefriction between the fingertip of the user and the touchscreen is modulated via electrostatic forces. When an al-ternating voltage is applied to the conductive layer of a

∗ The first two authors contributed equally to this work.1College of Engineering, Robotics and Mechatronics Laboratory, Koc Uni-versity, 34450, Istanbul, Turkey# Corresponding Author, email: cbasdogan@ku.edu.trAuthor Contributions: Y.V, M.K.S, and C.B designed research, Y.V. and A.Iperformed research and analyzed data, Y.V, A.I. and C.B wrote the paper.

surface capacitive touch screen, an attractive electrostaticforce is generated in normal direction between the fingerand the surface. By controlling the amplitude, frequency, andwaveform of the input voltage, the frictional force betweenthe moving finger and the touch screen can be modulated[2], [3]. This technology has a great potential especiallyin mobile applications including communication, games,education, data visualization, and interface development forblinds.

Although it is straightforward to generate a tactile stimuluson a touch screen using electrovibration, its effect on ourhaptic perception is poorly understood yet. One of the areasthat require further research in this regard is the realis-tic rendering of textures. In this study, we investigate theroughness perception of periodic virtual textures displayedby electrovibration. Our roughness perception of real textureshas been already investigated extensively in the literature [4]–[14]. Based on the multi dimensional scaling (MDS) studyconducted by Hollins et. al. [15], there are two main percep-tual dimensions in texture perception: roughness/smoothnessand hardness/softness. This finding was further confirmed byHollins et al. [16] and Yoshioka et al. [17] in later studies.Since it is not possible to alter softness on touch screens viaelectrostatic attraction yet, we have focused on roughnessperception.

In the literature, several types of stimuli have been usedto investigate roughness perception of real textures; raiseddots with controlled height and density [9], [12], ditheredcylindrical raised elements [4], [5], [11], and metal plateswith linear gratings [6], [10], [13]. These studies showedthat size of the tactile elements (i.e gratings, dots, cones) andthe spacing between them are critical parameters in rough-ness perception. Moreover, Hollins et al. [5] and Klatzkyand Lederman [18] found that the underlying mechanismbehind roughness perception is different for micro-textures(textures having inter-element spacing <≈ 0.2 mm) andmacro-textures (textures having inter-element spacing >≈0.2 mm).

At macro-textural scale, Lederman and collegues [4], [6],[10], [13] observed that groove width has a greater effect onperceived roughness than ridge width. This observation hasbeen supported by other studies later [14] suggesting thatthe perceived roughness increases monotonically with thegroove width, but decreases modestly with the ridge width.On the other hand, further increasing the spacing betweenthe tactile elements (more than 3.5 mm) causes the subjectsperceive the surface as smooth rather than rough. Hence,roughness perception follows an inverted U-shaped function

263

of groove width, reaching a maximum value at approximately3.5 mm of bump separation. Based on these observations,Lederman and Taylor developed a mechanical model thatestimated the perceived roughness as a power function ofthe total area of the skin that was instantaneously deformedthrough contact with the surface texture [4], [6]. Later,Connor et al. [12] developed a neural model which statedthat roughness perception of macro textures is achieved byspatial cues through pressure change and finger deformation.These arguments were later supported by showing that speedhas little effect on perceived roughness [9]. In contrast tothese studies, Cascio and Sathian [14] showed that rough-ness perception is affected by temporal frequency, which isdefined as the ratio of finger speed to wavelength of texture.They conducted experiments with different finger speeds onreal textures. Their results showed that roughness perceptionof the subjects increased as a function of temporal frequencyfor textures with varying ridge widths. Moreover, Smithet. al. [11] reported that roughness perception is positivelycorrelated with the rate of change in the lateral force. Thisresult also implies that the temporal cues play a role inroughness perception of macro-textures.

At micro-textural scale, Hollins and colleagues [5] re-ported that the roughness perception is mainly achievedby vibratory cues. Moreover, they found direct evidencethat roughness perception of micro-textures is mediated byPacinian Corpuscles. They used a Hall effect transducerto record the vibrations generated on the skin when a setof micro-textured surfaces is passively presented to theindex finger of subjects. They weighted the power of themeasured vibrations by the spectral sensitivity of Pacinianmechanoreceptors to show that roughness estimates of thesubjects were correlated with these weighted skin vibrations[19].

In this study, we have chosen our tactile stimuli at macro-textural scale. Since the tactile stimuli displayed to thesubjects in our study is independent of spatial position, ourstudy is best comparable to those on linear gratings. Inthose studies, the roughness perception has been typicallyinvestigated by square gratings of varying groove and ridgewidths. Due to the time-dependent nature of our signals, wehave opted to use ”wavelength” and ”duty cycle” as our

Input Voltage OutputElectrostatic Force

(a)

(b)

Fig. 1: The electrostatic forces generated by the model in[2]; a. without amplitude modulation, b. with amplitudemodulation. Amplitude modulation prevents the distortionsdue to high pass filtering of human electrical properties.

design parameters, but adjusted their values to match thegroove width values used in the literature while keepingthe ridge width constant. The effect of wave amplitudeon roughness perception has not been investigated in thisstudy since it is already known that perceived roughness ispositively correlated with the height of the surface features[20], [21].

This study contributes to the literature by answering thefollowing research questions: 1) How does our haptic percep-tion of roughness displayed by electrovibration change withwaveform and groove width? 2) How does our roughnessperception correlate with the contact forces acting on finger?

II. MATERIALS AND METHODS

A. Participants

We conducted an experimental study with 10 humansubjects (3 women, 7 men) to investigate the roughnessperception of haptic gratings. The ages of the subjects werevaried between 22 and 30 (average = 26.4 ± 2.8). Thesubjects used the index finger of their dominant hand toexplore the surface. They washed their hands with soap andrinsed with water before the experiment. Also, their fingersand the touch screen were cleaned by alcohol before eachmeasurement. The subjects read, and then signed the consentform before the experiments. The form was approved byEthical Committee for Human Participants of Koc University.

B. Stimuli

We used the concept of amplitude modulation to generateour electrovibration stimuli. We modulated the amplitude ofa low-frequency envelope signal, fe, with a high frequencysinusoidal carrier signal, fc, at 3kHz. Since 3kHz is far be-yond the human vibration sensation range (up to 1kHz), thiscarrier signal did not cause any additional vibratory feelings.Moreover, the proposed amplitude modulation minimizesthe perceptional differences due to the electrical filteringof human stratum corneum and eliminates low-frequencydistortions as shown in Fig. 1 [2]. The finger scan speed,v, was maintained at approximately 50 mm/s during theexperiments using a visual cursor displayed to the subjects.We selected this scan speed based on the values used in theliterature [22].

In our implementation, we only consider the positive partsof the modulated signal (i.e. unipolar signal), y(t), to displayvirtual textures on the touch screen. Due to the fact that theelectrostatic force is a function of the square of the inputvoltage signal, a unipolar input (voltage) to the touch screenresults in a unipolar output (electrostatic force) with thesame wavelength, λ , and duty cycle of the input (Fig.1).An illustration of stimuli generation in our system is shownin Fig. 2. In addition to using a sine wave, we also utilizedsquare, saw-tooth, and triangle waves as the modulated signaland altered their wavelength and duty cycle to generatedifferent virtual textures.

We evaluated the roughness perception of 40 differentvirtual textures (Table I). The voltage amplitude for eachstimulus was 100 V. We set the wavelengths and duty cycles

264

Saturation

Displayed Signal Resultant ForceWavelength Amplitude

Pulse Width

Duty Cycle = Pulse Width/ Wavelength

Fig. 2: Stimuli generation via electrovibration. A low frequency unipolar signal is modulated with a high frequency (3kHz)carrier signal. Only the positive parts of the signals are considered as the input to the touch screen. Due to the high frequencymodulation, the resultant electrostatic force has the same wavelength and duty cycle as the input voltage signal.

of our virtual textures to match the typical groove widthvalues used in the literature (Table I). The ridge width waskept constant as 0.5 mm. The resultant temporal frequency,fe, was derived from v

λ.

TABLE I: Experimental parameters and their correspondingvalues. (GW: Groove Width, RW: Ridge Width)

Controlled Parameters Corresponding Parameters

WaveformGW

(mm)RW

(mm)Wavelength

(mm)DutyCycle

TemporalFrequency

(Hz)

Square

Sinusoidal

Triangular

Saw-tooth

0.125

0.5

0.625 0.8 80

0.250 0.750 0.6666 66.67

0.375 0.875 0.5714 57.14

0.5 1 0.5 50

1 1.5 0.3333 33.33

1.5 2 0.25 25

2 2.5 0.2 20

3.5 4 0.125 12.5

5.5 6 0.0833 8.33

7.5 8 0.0625 6.25

Each texture was displayed 6 times to the subjects. Hence,the total number of haptic stimuli for each subject was 240(40 texture × 6 repetitions). The stimuli were displayed inrandom order while changing the order for each subject ineach session.

C. Experimental Setup

A surface capacitive touch screen (SCT3250, 3M Inc.) wasused to display tactile stimuli. Desired signals were generatedby a DAQ card (PCI-6025E, National Instruments Inc.) andboosted by an amplifier (E-413, PI Inc.) before transmittedto the touch screen. A compact monitor displaying a visualcursor moving with a speed of 50 mm/s was placed belowthe touch screen to adjust the scan speed of the subjects (Fig.3). A force sensor (Nano17, ATI Inc.) was placed under thetouch screen to acquire normal and tangential forces duringthe experiments. Both the DAQ card and the force sensorhad a sampling rate of 10kHz. Also, an IR frame was placedabove the touch screen to detect the actual scan speed of thesubjects. This information was used to check if the actualscan speed of a subject had been significantly different fromthe desired value in each trial. If so, all the related data ofthat trial was discarded to obtain more consistent results.

100 mm

Movingcursor

Touch screen

40 m

m

Forcesensor

Fig. 3: The haptic exploration area on the touch screen.

D. Experimental Procedure

The subjects were instructed to sit on a chair in front ofthe setup and move their index finger back and forth twicein the horizontal direction to explore the haptic stimuli.

During the experiments, the subjects were allowed toreplay the haptic stimuli only once by pressing the replay but-ton. To reduce tiring of the subjects, an arm rest was placedunder their wrist. The subjects were asked to wear noisecancellation headphones to prevent from any noise affectingtheir haptic perception. Before starting the experiment, thesubjects were given instructions about the experiment andpresented with a training session displaying all stimuli oncein random order. It took about 30 minutes for each subjectto complete one session including training. The experimentswere performed in three separate sessions.

A Likert scale of 1-7 was used to rate the haptic perceptionof the subjects. After each trial, the subjects entered theirrating of the stimulus using a pen-based stylus and a simplegraphical user interface consisting of 7 radio buttons.

E. Data Analysis

The roughness estimates of each subject were normalizedbetween 0 and 1 after each experimental session. Then, foreach texture, the average of six estimates were calculated.

During each trial, both normal and tangential forces wererecorded using the force sensor. A data segment of onesecond was chosen from the last stroke of each trial andthe following metrics were calculated: rms of tangentialforce (RTF), rms of normal force (RNF), average frictioncoefficient (AFC), rms of rate of change in the tangentialforce (RTFR) and rms of rate of change in the normal force(RNFR). AFC was calculated by dividing the average tangen-tial force to the average normal force. RTFR and RNFR were

265

calculated as backward differentiation of recorded tangentialand normal force, respectively. Each metric was normalizedbetween 0 and 1 for each subject and session. Then, theaverage of six measurements was reported for each texture.

III. RESULTS

The normalized roughness estimates (means and standarderrors) as a function of duty cycle, wavelength, and temporalfrequency are shown in Fig. 4. Since the values for dutycycle, wavelength, and temporal frequency were selected toalter groove width (while keeping the ridge width constant),duty cycle and wavelength -or duty cycle and frequency-were coupled with each other. Yet, they were two differentdimensions. The results were analyzed using 2-way ANOVArepeated measures. The results showed that both waveformand groove width had significant effects on the roughnessperception of the subjects (Table II). Also, there was astatistically significant interaction between waveform andgroove width. Bonferroni corrected paired t-tests showedthat the subjects perceived the square wave as the roughest(p<0.01), while the other waveforms were perceived equallyrough. Moreover, the difference in roughness perception wassignificant for the groove width varying between 1.5 and 7.5mm (p<0.01), but not significant for the ones below 1.5mm. Greenhouse-Geisser method was used to correct thefractional degrees of freedom in F-value.

TABLE II: ANOVA results for roughness perception.

Effects F-value p-value

Waveform F(2.01,18.09) = 14.137 < 0.01∗

Groove Width F(9,81) = 3.289 < 0.01∗

Waveform × Groove Width F(27,243) = 1.811 0.01∗

The measured values (RNF, RTF, AFC, RNFR, RTFR) asa function of groove width are shown in Fig. 6. We alsoanalyzed these results using 2-way ANOVA repeated mea-sures. The results showed that the main effects (waveformand groove width) and their interaction had no significanteffect on RTF, RNF and AFC, but on RTFR and RNFR.Bonferroni corrected paired t-tests showed that RTFR andRNFR were significantly different for all waveform pairs(p<0.01) except triangular and saw-tooth waves. Moreover,the difference in RTFR and RNFR were significant for thegroove width values between 1 mm and 7.5 mm (p< 0.01),but not significant for the ones below 1 mm.

TABLE III: ANOVA results for force metrics.

MetricEffects

Waveform Groove Width Waveform x Groove Width

RNF p = 0.310 p = 0.100 p = 0.102

RTF p = 0.078 p = 0.100 p = 0.200

AFC p = 0.340 p = 0.700 p = 0.279

RNFR p <0.01∗ p <0.01∗ p <0.01∗

RTFR p <0.01∗ p <0.01∗ p <0.01∗

IV. DISCUSSION

Our results showed that the subjects perceived the squarewave as the roughest while the other waveforms (sinusoidal,triangle and saw-tooth) were perceived equally rough (Fig.4). Hence, it can be deduced that our perception of roughnessdepends on the waveform of the tactile stimulus displayed byelectrovibration. The earlier studies also support our findings.In our previous study [2], we measured the tactile detectionthresholds of sinusoidal and square wave signals displayed byelectrovibration. We found that square wave signals are moredetectable than sinusoidal waves at frequencies lower than60 Hz. Cholewiak et al. [23] conducted human experimentsin virtual environments using a haptic device to investigatethe amplitude detection thresholds for virtual gratings in theform of sinusoidal and square waves. The results showedthat the detection threshold for the square-wave gratingswas lower than that of the sinusoidal gratings. Kocsis et al.[24] conducted discrimination experiments with both real andvirtual sinusoidal and triangular textured surface gratings. All

0.2

0.3

0.4

0.5

0.6

0.70.80.9

1

10-1

101100 100

squ

Rou

ghne

ss E

stim

ates

(Nor

mal

ized

)

Duty CycleWave Length (mm)

sin

(a)

0.2

0.3

0.4

0.5

0.6

0.70.80.9

1

10-1

102100 101

saw

Frequency (Hz)

tri

(b)

Fig. 4: Roughness estimates (means and standard errors) of10 subjects as a function of duty cycle, wavelength, andtemporal frequency.

1000.2

0.3

0.4

0.5

0.6

0.70.80.9

1

Groove Width (mm)

Rou

ghne

ss E

stim

ates

(N

orm

aliz

ed)

Groove Width (mm)10 0 10

0.2

0.3

0.4

0.5

0.6

0.70.80.9

1

[15] (Real RW = 0.5 mm)[15] (Real RW = 0.5 mm)[15] (Real RW = 0.5 mm)[15] (Real RW = 0.5 mm)[28] (Virtual Probe, r = 0.5 mm)[28] (Virtual Probe, r = 1.5 mm)[28] (Virtual Probe, r = 0.25 mm)[22] (Virtual RW = 0.625 mm)

[28] (Virtual Probe, r = 1 mm )

(Virtual, RW = 0.5 mm)

sinsqu

sawtri

(a) (b)

840.2 0.440.2 0.4

Fig. 5: a. Roughness estimates (means and standard errors)of 10 subjects as a function of groove width. b. Roughnessestimates reported in literature.

266

100

Groove Width (mm)

0.2

0.4

0.6

0.81

RN

F (N

orm

aliz

ed)

1000.2

0.4

0.6

0.81

RTF

(Nor

mal

ized

)

10010-1

1

AFC

(Nor

mal

ized

)

100

10-2

10-1

100

RN

FR (N

orm

aliz

ed)

100

10-2

10-1

100

RTF

R (N

orm

aliz

ed)

Groove Width (mm) Groove Width (mm) Groove Width (mm) Groove Width (mm)

squ sin saw triRidge Width = 0.5 mm

(a) (b) (c) (d) (e)

0.2 8 0.2 80.2 8

0.6

0.4

0.2

0.2 8 0.2 8

Fig. 6: Normalized force metrics (means and standard errors) a. RMS of normal force b. RMS of tangential force c. Averagefriction coefficient d. RMS of rate of change in normal force e. RMS of rate of change in tangential force.

gratings had a wave length of 2.5 mm while their height(wave amplitude) varied between 55-70 µm. The resultsshowed that the discrimination thresholds did not differsignificantly between sinusoidal and triangular gratings.

We found that the relation between roughness perceptionand groove width follows an inverted U-shaped curve asreported in earlier studies (Fig. 5). In those studies [13], [20],[25], the roughness perception reached its maximum value atgroove width of approximately 3.5 mm, but it was 1.5 mmin our study. This discrepancy can be attributed to severalfactors. Hollins et al. [5] and Klatzky and Lederman [18]stated that the perception of real micro-textures is mainlyachived by vibratory cues via Pacinian Corpuscle mechanore-ceptors while that of the macro-textures is achieved by spatialcues through finger deformation and change in pressure.Although the textures were at macroscale in our study, thehaptic stimuli were mainly perceived through the vibrotactilechannel and the influences of pressure change and fingerdeformation on this perception were less significant (seeFig. 7). Another possible reason for the difference couldbe the effect of finger speed on the haptic perception oftextures in our study. In the earlier studies performed withbare finger and real textures [10], the effect of speed ontexture perception has found to be negligible, suggestingthat the roughness perception is mainly governed by spatialcues. However, when the effect of spatial cues is reducedsignificantly and the vibration is the only feedback as inthe case of exploring textures with a rigid probe, the speed

Fig. 7: Comparison of normal forces applied to fingertipwhile touching real textures and the virtual ones displayedby electrovibration on touch screens. Here, Fn is the appliednormal force (quasi-constant during exploration) while Fe isthe time-dependent electrostatic force.

of exploration appears to affect the haptic perception [4].In fact, the location of the peak point shifts in roughnessversus groove width curve with respect to the explorationspeed. Cascio and Sathian [14] also showed that roughnessperception is affected by temporal frequency, which dependson exploration speed.

Finally, our force measurements showed that RTFR andRNFR are the ones best correlated with the roughnessperception of subjects. To better understand the reasoningbehind this, we estimated the rms of electrostatic force (REF)and its derivative (REFR) using the model in [2] for thewaveform and groove width values utilized in our experiment(Fig. 8). Here, REF and REFR were the highest for the squarewave and lowest for triangular and saw-tooth waves. REFdecreased with the increase in groove width in our case,since the corresponding duty cycle values decreased (seeTable I). In addition to the reduction in the force magnitude,the decline in the temporal frequency as the grove widthincreases caused a similar trend in REFR. However, thedifference caused by electrostatic force did not appear on themeasured force magnitudes but on their derivatives. Since thenormal forces applied by the subjects were not controlledin our experiment, the resultant force magnitudes (normaland tangential) showed fluctuations though they reduced asa function of increasing groove width (Figs. 6a and 6b),as observed in the simulations (Fig. 8a). Moreover, the

100

Groove Width (mm)

10-3

10-2

10-1

100

REF

(Nor

mal

ized

)

100

Groove Width (mm)

10-3

10-2

10-1

100

REF

R (N

orm

aliz

ed)

squ sin saw tri

(a) (b)

Ridge Width = 0.5 mm

0.2 8 0.2 8

Fig. 8: a. RMS of the electrostatic force (REF) and its rateof change (REFR) calculated using the model in [2].

267

electrostatic force modulated the tangential force accordingto the Coulomb model, which resulted in an approximatelyconstant friction coefficient as suggested in [3] and shown inFig. 6c. In contrast to the real textures, the time-dependentchange in the contact forces depends on only electrostaticforces which are almost ten times smaller than the appliedfinger force. This is why RTFR and RNFR were bettercorrelated with the roughness perception of subjects, andtheir trends as a function of groove width were matched tothat of the simulations (compare Figs. 6d, 6e with Fig. 8b).

V. CONCLUSIONIn this study, we conducted psychophysical experiments

with 10 subjects to investigate the human roughness per-ception on electrostatic displays. We rendered sinusoidal,square, saw-tooth, and triangular pulse waves in differentgroove widths while keeping the ridge width constant. Wemodulated these low frequency pulse waves with a 3kHzhigh frequency signal to prevent low frequency distortionson the electrostatic force. Our results showed that humanroughness perception follows an inverted U-shaped trendwith respect to groove width, as reported in the earlierstudies. The subjects perceived square wave as the roughestwhile their haptic perception of the other waveforms wasnot significantly different from each other. Moreover, theroughness perception of the subjects was best correlated tothe rate of change of the normal and tangential forces ratherthan their magnitudes.

To our knowledge, this is the first detailed study in-vestigating human roughness perception of virtual textureson a touch screen. In addition to providing an insight toour understanding of human haptic perception, engineersand designers can benefit from our study while developingapplications which involve rendering tactile effects on atouch screen. They can adjust wavelength, duty cycle andwaveform of the voltage signals to generate tactile effectswith desired roughness.

In this study, we adjusted the wavelength and duty cycle ofour signals to match the groove width values in the literature.This gave us the opportunity to compare our results withthose of the earlier studies. However, this choice did notallow us to investigate the individual effects of wavelengthand duty cycle on our roughness perception, which we planto do so in the near future. We will also expand our studyby including micro-textures.

ACKNOWLEDGEMENTSThe Scientific and Technological Research Council of

Turkey supported this work under BIDEB-2211 Program.

REFERENCES

[1] O. Bau, I. Poupyrev, A. Israr, and C. Harrison, “Teslatouch: Elec-trovibration for touch surfaces,” in Proc. of the 23nd annual ACMsymposium on User interface software and technology (UIST’10),NewYork, USA, 2010, pp. 283–292.

[2] Y. Vardar, B. Guclu, and C. Basdogan, “Effect of waveform inhaptic perception of electrovibration on touchscreens,” in Haptics:Perception, Devices, Control, and Applications: 10th InternationalConference, EuroHaptics 2016, London, UK, July 4-7, 2016, Proceed-ings, Part I, 2016.

[3] D. Mayer, M. Peshkin, and E. Colgate, “Fingertip electrostatic mod-ulation due to electrostatic attraction,” in Proc. IEEE World HapticsConference (WHC’13), Daejeon, South Korea, Apr. 2013, pp. 43–48.

[4] R. Klatzky, S. Lederman, C. Hamilton, M. Grindley, and R. Swendsen,“Feeling textures through a probe: Effects of probe and surface ge-ometry and exploratory factors,” Perception & Psychophysics, vol. 65,pp. 613 – 631, 2003.

[5] M. Hollins, S. Bensmaia, and S. Washburn, “Vibrotactile adaptationimpairs discrimination of fine, but not coarse, textures,” Somatosensoryand Motor Research, vol. 18, no. 4, pp. 253–262, 2001.

[6] S. Lederman and M. Taylor, “Fingertip force, surface geometry,and the perception of roughness by active touch,” Perception &Psychophysics, vol. 12, pp. 401 – 408, 1972.

[7] S. Lederman, G. Thorne, and B. Jones, “Perception of texture byvision and touch: Multidimensionality and intersensory integration,”J. Experimental Psychology: Human Perception and Performance,vol. 12, pp. 169–180, 1972.

[8] C. Connor and K. Johnson, “Neural coding of tactile texture: Compar-ison of spatial and temporal mechanisms for roughness perception,”J. Neuroscience, vol. 12, pp. 3414–3426, 1992.

[9] S. Lederman, R. Klatzky, C. Hamilton, and G. Ramsay, “Perceivingroughness via a rigid probe: Psychophysical effects of explorationspeed and mode of touch,” Haptics-e: The Electronic J. HapticsResearch, vol. 1, pp. 1–20, 1999.

[10] S. Lederman, “Tactile roughness of grooved surfaces: The touchingprocess and effects of macro- and microsurface structure,” Perception& Psychophysics, vol. 16, no. 2, p. 395, 1975.

[11] A. M. Smith, C. E. Chapman, M. Deslandes, J. S. Langlais, andM. P. Thibodeau, “Role of friction and tangential force variation in thesubjective scaling of tactile roughness,” Experimental Brain Research,vol. 144, no. 2, pp. 211–223, 2002.

[12] C. Connor, S. Hsiao, J. Philips, and K. Johnson, “Tactile roughness:Neural codes that account for pscyhophysical magnitude estimates,”The journal of neuroscience, vol. 10, pp. 3823–3836, 1990.

[13] M. Lawrance, R. Kitada, R. Klatzky, and S. Lederman, “Hapticroughness perception of linear gratings via bare finger or rigid probe,”Perception, vol. 36, pp. 547–557, 2007.

[14] C. Cascio and K. Sathian, “Temporal cues contribute to tactile per-ception of roughness,” Journal of Neuroscience, vol. 21, no. 14, pp.5289–5296, 2001.

[15] M. Hollins, R. Faldowski, R. Rao, and F. Young, “Perceptual dimen-sions of tactile surfaced texture: A multidimensional scaling analysis,”Perception & Psychophysics, vol. 54, no. 6, pp. 697–705, 1993.

[16] M. Hollins, S. Bensmaia, K. Karlof, and F. Young, “Individualdifferences in perceptual space for tactile textures: Evidence frommultidimensional scaling,” Perception & Psychophysics, vol. 62, no. 8,pp. 1534–1544, 2000.

[17] T. Yoshioka, S. Bensmaia, J. Craig, and S. Hsiao, “Texture perceptionthrough direct and indirect touch: An analysis of perceptual spacefor tactile textures in two modes of exploration,” Somatosensory andMotor Research, vol. 24, no. 1-2, pp. 53–70, 2007.

[18] R. L. Klatzky and S. J. Lederman, Multisensory Texture Perception.New York, NY: Springer New York, 2010, pp. 211–230.

[19] S. Bensmaia and M. Hollins, “Pacinian representations of fine surfacetexture,” Percept Pscyhophys, vol. 67, pp. 842–854, 2005.

[20] R. Klatzky and S. Lederman, “The perceived roughness of resistivevirtual textures: I. rendering by a force-feedback mouse,” ACM Trans.Applied Perception, vol. 3, no. 1, pp. 1–14, 2006.

[21] D. Blake, K. Johnson, and S. Hsiao, “Neural coding mechanismsin tactile pattern recognition: The relative contributions of slowlyand rapidly adapting mechanoreceptors to perceived roughness,” Thejournal of neuroscience, vol. 17, pp. 7480–7489, 1997.

[22] T. Yoshioka, J. Craig, G. Beck, and S. Hsiao, “Perceptual constancy oftexture roughness in the tactile system,” The Journal of Neuroscience,vol. 31, no. 48, p. 1760317611, 2011.

[23] S. A. Cholewiak, K. Kim, H. Z. Tan, and B. D. Adelstein, “A frequencydomain analysis of haptic gratings,” IEEE Transactions on Haptics,vol. 3, no. 1, pp. 3–14, 2010.

[24] M. B. Kocsis, S. A. Cholewiak, R. M. Traylor, B. D. Adelstein,E. Daniel Hirleman, and H. Z. Tan, “Discrimination of real and virtualsurfaces with sinusoidal and triangular gratings using the fingertip andstylus,” IEEE Transactions on Haptics, vol. 6, no. 2, pp. 181–192,2013.

[25] B. Unger, R. Hollis, and R. Klatzky, “Roughness pereption in virtualtextures,” IEEE Transactions on Haptics, vol. 4, 2011.

268